Does the political regime affect disaster vulnerability?
A thesis presented to
the Faculty of Global Affairs
Leiden University
In partial fulfillment of the requirements
for the Degree of Master of Science
By
Nick Kempers
S1560603
June 2020
Table of contents
Introduction 3
Chapter 1: Theoretical framework 7
Political regimes & disaster management 7
Vulnerability & disaster management 10
Vulnerability & political regimes 11
Pressure and Release model 12
Conceptualization 15 Hypothesis 16 Chapter 2: Methodology 19 Research design 19 Operationalization 20 Case selection 24 Data analysis 25 Chapter 3: Analysis 27 Empirical results 27 Test of Normality 30 Pearson Correlation 31 Linear regression 33 Statistical analysis 35
Applying the PAR-model 37
Results 40 Discussion 42 Limitations 44 Implications 45 Conclusion 46 Bibliography 48
Appendix
Appendix I: The complete set of variables for PVI and its indicators Appendix II: Data for the PVI-ES variables
Appendix III: Z-Scores for the PVI-ES variables Appendix IV: Data for the PVI-SF variables Appendix V: Z-scores for the PVI-SF variables Appendix VI: Data for the PVI-LR variables Appendix VII: Z-scores for the PVI-LR variables
Appendix VIII: Scatterplot for the Democracy Index and PVI-ES Appendix IX: Scatterplot for the Democracy Index and PVI-SF Appendix X: Scatterplot for the Democracy Index and PVI-LR
Introduction
Disasters do not hit states equally hard. Not only because of a natural indifference but also due to a multitude of social, economic, and political factors that influence how much damage a disaster can do. We can illustrate this fact by comparing two hurricanes that were similar in power but which hit different countries: hurricane Katrina & cyclone Nargis. Hurricane Katrina hit the United States in 2005 with winds up to 280 kilometers per hour and killed around 1.200 people directly as well as costing 108 billion dollars in damages. After this disaster, a discussion started about how the hardest-hit city of New Orleans could have been as vulnerable as it was (Gibbens, 2019). Less than three years later, a storm which was a bit weaker made landfall in Myanmar. This cyclone ravaged through the autocratic state and killed almost 140.000 people, while the damages were around 10 billion dollars. In the aftermath, the military junta, which rules the country, was mainly criticized for not accepting international aid, which would have led to many unnecessary deaths (Fritz, Blount, Thwin, Thu & Chan, 2009). While this comparison only scratches the surface of the suffering that these cyclones created, it already shows significant differences between the effects of two similar disasters and their government’s responses.
Is it possible that there were fewer deaths in the United States than there were in Myanmar because they are a democracy while the other is an autocracy? If so, does this difference exist on a global scale, which could identify the vulnerability of individual states? This thesis will explore this question by looking into how vulnerable different political regimes are and looking for a relationship between these two variables. If this relationship exists, an explanation will be sought to explain why some regime types are less vulnerable than the others are.
That these political regimes could differ in how vulnerable they are can be expected when looking into earlier findings in the field of crisis management. Strömberg (2007) found out that democracies seem to suffer more fatalities and wounded than autocracies do, which he thinks is counterintuitive. He would have thought that democracies have a better state capacity, which means that they can react faster and more adequately to a disaster than other states can. Diamond & Linz (1989) found that being democratic has another benefit, as they are less vulnerable to economic downfalls. In autocracies, these economic hardships have a chance to undermine the legitimacy of the ruling elite as they have fewer resources to keep them happy. Being an
autocracy is not only disadvantageous, however, as Omelicheva (2011) used a statistical analysis that discovered that autocratic states that suffered a natural disaster are the least likely to
destabilize into conflict. In contrast, regimes that are in transit or are democratic and semi-autocratic are most likely to fall. Lastly, there even are differences within the regimes as Landry & Stockmann (2009) found that one party-systems are the most resilient of all autocracies. All of these findings combined show us that in disaster management, there are apparent differences between different political regimes, which we use as the starting point of the argument.
These findings do not tell if there is a relationship between disaster vulnerability and the political regime and how it would look. To be able to find this relationship, these concepts first need to be defined, and this is already a challenge for vulnerability. Currently, there are over 25 different definitions within the literature, which shows how multifaceted it is but also that it is still not exactly clear what it entails. The concept arose to challenge the hazard-oriented view, which was predominant within disaster management in the 80s, and it combined how susceptible people were with their abilities to deal with the damage that could occur (Birkmann, 2006). In 2004, the most common definition was written by the United Nations International Strategy for Disaster Reduction, which is still the same at its core: ‘The conditions determined by physical, social, economic and environmental factors or processes, which increase the susceptibility of a community to the impact of hazards’ (United Nations International Strategy for Disaster
Reduction, 2004). As this definition is too general, a clear definition and system need to be set up for this thesis to use this concept, for which the Prevalent Vulnerability Index is used (Cardona, 2005). The other variable, the political regime, is measured by using the Democracy Index of the Economists Business Intelligence Unit (2015). This yearly index measures the amounts of
freedoms the citizens of a country enjoy and how repressive their current government is. This paper will also use the four categories of political regimes in which they sorted them.
With these concepts, this paper then seeks out to find if this relationship between vulnerability and political regime exists. For this, the following research question needs to be answered: ‘To
what extent does the type of regime affect its disaster vulnerability?’ This thesis will select
twelve states based on a most different system design, which are equally divided over the four regime categories that are given by the Democracy Index. For these states, the data for the variables of the Prevalent Vulnerability Index needs to be gathered. This data comes mostly from the different databases that are provided by the World Bank. The complete index will then be analyzed by using different statistical methods, as the Pearson correlation, to find out if there is a significant relationship between how democratic a state is and how vulnerable it is. The (lack of) relationship will then be explained by using the Pressure and Release-model, as described by Wisner, Blaikie, Cannon & Davis (2004). This model gives us the tools to identify where the cause of a particular vulnerability lies. It describes that there are three layers, of which the first one is the political and the economic roots of the state. The second one is the dynamic pressures, societal or demographic changes that can signal vulnerabilities as these changes have to be influenced by an external event (e.g., an economic recession). The bottom layer is the unsafe conditions, which are situations that citizens directly deal with. These conditions could be deferred maintenance on critical infrastructure, which makes it more likely to collapse and thus lead to disaster. This framework then is used to identify in which level the variables of the Prevalent Vulnerability Index fit, and if these possibly have a political root.
The findings of this paper can help researchers understand how the institutions and political ideology influence or predict their vulnerability towards disasters. By identifying internal processes and structures that are responsible for these differences, academics and policymakers can try to solve these structural weaknesses for a type of political regime. In practice, this would mean that if one institution in autocratic systems is generally more vulnerable to disasters than they are in democracies, policymakers can try to find ways to mitigate this inherent vulnerability. As a result, more people worldwide can be better protected against natural disasters. More
others, which can be used by (non-governmental) international organizations. These
organizations can then use this information about possible vulnerabilities to prepare accordingly for future disasters. Lastly, if it is true that one political regime may be better suited to withstand disasters than others, this can be used for the promotion of this regime type on the world stage.
Next to the societal importance, this thesis has value for the academic field of disaster management. It would help fill a gap in research as authoritarian states are structurally under-researched in the field of crisis management. Most studies focused on (Western) democracies, which means that their findings are hardly representative of the whole world (Chan, 2013). Around half of all countries are currently classified as at least non-democratic (Freedom House, 2020, p. 20), which shows that this group needs to be considered when you want to generalize the findings of your research. This thesis will strive to show that there is not only a difference between autocracies, democracies, and the regimes that fall in between but that there is also regularity in this difference between them. When this succeeds, it could add to a small set of comparative literature regarding political regimes and disasters.
This paper consists of four chapters that will guide the reader through the research. The first chapter is the theoretical framework, which explores earlier findings of the following main concepts: political regimes, disaster management, and vulnerability. It also discusses how they relate to each other to create a clear overview of which information is most important. The second chapter is the methodology that describes the exact steps on how to replicate this
research. Critical decisions such as the conceptualization, operationalization, case selection, and data collection are all thoroughly discussed so the reader understands which choices were made and with which reasons. The third chapter is the analysis, and it discusses the results of the quantitative research together with the qualitative analysis based on the knowledge of the theoretical framework. The last part is the discussion, which summarizes all the knowledge and findings of this thesis. It also acknowledges some limitations and makes recommendations for future research.
Chapter 1: Theoretical framework
To answer the question to what extent the political regime may influence its disaster
vulnerability, we need to assess what has already been written regarding this topic. This chapter will explore the current literature on the main concepts of this paper. The first one is the regime type, or political regime, and what we currently know about how it relates to the academic field of disaster management. The second concept is vulnerability of the initial focus on why it is an essential idea within disaster management. Lastly, we explore academic literature that has combined these two concepts to find if there are already some (in)direct relationships. This prior knowledge can then can be used for building a hypothesis. Explaining what this hypothesis is and on what assumptions it is based ends this first chapter.
Political regimes & disaster management
As mentioned before, this chapter starts by explaining the main concepts separately to create a broader overview of the current literature. This facilitates a better understanding of them, as well as the relationships between them and disaster management. In this field, the effect of the regime type has already been researched extensively. The following examples show that it is already accepted that there is a clear division between democratic and non-democratic nations and that these differences are not accidental.
The finding that stands out the most comes from Strömberg in his research on how disasters, economic development, and humanitarian aid are related. Using the EM-DAT database, which is a database that collects all disasters and damages that strike around the globe, he qualitatively analyzed if disasters became more frequent, which countries suffer the most, and he explains these findings. When looking into how many disasters there are, how many people they killed, and how many were affected, he finds that democratic nations are more likely to suffer more fatalities. He is hesitant to conclude this, however, as he thinks that democracies tend to report more accurately than other regimes. This would negatively influence the data for democracies. (Strömberg, 2007, p. 209).
Another consequence of disasters is its economic damage, and autocracies seem to be harder hit than democracies are. Gassebner, Keck & Ter (2006) used EM-DAT data to compare states which are hit by a disaster and how this influences their imports and exports. One of the
variables that they found that influence how the trade of these states is affected after a disaster is how democratic it was: the less democratic a state is, the more trade they lost. The consequences of the disaster are not the only thing that differs between regimes, but the actions the
governments take also do. Howard, Agarwal & Hussain (2011) applied an event history analysis and looked for news articles describing significant disruptions in the digital infrastructure. They then categorized them into regime types and the reason for which it happened and found that authoritarian regimes most often interfere in their national digital environment. While the reasons did differ, it was also often done to limit the spread of information, making sure that the government's story can not be challenged. Another clear example of how the actions of various regimes differ is the use of language by political leaders. An analysis of political speeches showed that autocratic leaders are more likely to use blaming and credit-claiming in their post-disaster speeches than their democratic counterparts are (Windsor, Dowell & Graesser, 2014).
Until now, democratic and autocratic nations were the focus of this theoretical section, but some countries are neither. Ahlerup (2013) focuses on regimes that were in a transition towards a democracy and looked if the democracy score of a state significantly rose when it got
humanitarian aid. He found out that natural disasters accelerate their democratization process as most aid donors are of a democratic nature (e.g., countries or international organizations). In exchange for aid, they often require the disaster-stricken state to make institutional changes that make them more democratic overall. Lastly, there is a group of countries which has combined autocratic and democratic institutions into a hybrid regime, which have characteristics from both democracies and autocracies. Omelicheva (2011) used a statistical method to determine which variables can predict if a state becomes politically unstable. How unstable a state is, is measured using the data of the Political Instability Task Force. This database keeps track of all conflicts worldwide, how big they are, and their effects. With this data and a logistic regression, she found that these hybrid regimes have the highest chance to collapse into political instability after being hit by a disaster.
These findings all show that differences between regime types in the field of disaster
management are common. Still, these are specific examples which lack a universal explanation or effect. A few academics have searched to explain these differences more generally, but a consensus has not been found yet. Chan (2013), for example, is most critical towards the field of disaster management, as it is dominated by theories that are crafted from experiences in wealthy, democratic countries. This focus negatively affects the usability of these theories when using them on an international scale. He then set out to explain why democratic states should be more resilient against disasters, using a combination of the state-societal framework and the crisis development ladder as his theoretical framework. He argues that democracies inherently have checks and balances which keep the government in check and that non-democracies lack these. Especially having chosen representatives and having free media are essential after a disaster, as it means that the public has access to different channels of information and can hold its
representatives responsible. In a democracy, this means that when a representative did not act adequately in the eyes of the population, he/she will be replaced democratically. In an autocratic system, this controlled change of leadership is not generally possible in its institutional
framework. Chan then argues that when the population knows the government censors the actual effects of a disaster, and this leads to heavy criticism. This censorship can become a government crisis if the population starts questioning the government's authority. Then, this societal pressure needs just a small catalysator to increase pressure to the point that it is too much for the
government. This way, they are forced to change some policies or even their leadership. This theory suggests that the heavy institutionalization of democracies made these more resilient and have built-in systems to deal with sudden change. Autocracies, which are more robust, are more vulnerable when their populations become dissatisfied with their rulers.
This section shows that there has been much writing about specific differences between the political regimes when dealing with disasters. It also shows that there is still a gap in the academic literature when looking to explain these differences on a macro-level. The following part will explore how the core concept of vulnerability fits within the field of disaster
Vulnerability & disaster management
The previous section showed that the political regime does influence multiple aspects of disaster management, but how does it relate to vulnerability? To find out, we first need to explore what vulnerability means. While it is one of the core concepts within disaster management, it is still widely discussed. The differences go from seeing it as purely a social construct to seeing it as something that needs to be measured structurally, but the predominant views always include the economic, social, and political effects of disasters. The vulnerability of an actor is often seen as the combination of its liabilities and capabilities, which entails as well physical as social
elements (McEntire, Crocker & Peters, 2010). Antofie, Doherty, & Marin-Ferrer (2018, p. 11) also argue that vulnerability commonly has four features that are named when described: it is multi-dimensional, dynamic, scale-dependent, and site-specific. Academics even measure it at two different moments, some look into the amount of damage done to a system after a hazard, while others look at the state of a system before it encounters a hazard.
These definitions hit the core of what vulnerability means: to which extent are people and assets likely to be affected by disasters. However, this definition is far too abstract to be used in this paper. This is why most researchers measure vulnerability in a way that best suits their field of interest. Economically, the vulnerability of a state might be measured in the expected total loss of economic value after a disaster, while in social science, it is more interesting to look into how many lives were lost and which social groups are more affected than others. Flanagan, Gregory, Hallisey, Heitgerd & Lewis (2011) wrote a paper to include the social vulnerability index into the commonly used disaster cycle, as they believe that the earlier mentioned ways of measuring vulnerability are insufficient in most cases. This Social Vulnerability Index includes four
domains: socioeconomic status, the composition of the households, the minority status, and their languages and how their citizens live and travel. This index is a clear example of how
vulnerability can be found in different aspects of people's lives and heavily depends on how the researcher wants to measure it. The methodology section of this thesis will determine how vulnerability will be measured and why it will be measured that way.
As stated before, vulnerability is a core issue in disaster management, which means that it has been a frequent subject for research. Academics have already identified many variables that are likely to influence the vulnerability of a state. Benson (2003, p.12) argues that poor and socially disadvantaged groups, which can be whole states or groups within a state, are the most
vulnerable to disasters. These disasters, in turn, make the poor more miserable, creating a downward cycle of persistent poverty. Another known trend is that small island states are far more vulnerable to disasters than other countries are, especially the small and lesser connected islands. This vulnerability mostly stems from their remoteness and size; as a consequence of this, they have limited disaster mitigation capacities (Pelling & Uitto, 2001).
Vulnerability & political regimes
In the last sections, we saw that the regime type of a state does matter in the field of disaster management, and that vulnerability is a broad but essential concept within this field. The link between these two, however, is less studied. Lin (2015) did do a study in which he studied which factors might influence the disaster mitigation of a state. By using EM-DAT data from 1995 to 2005, he analyzed the deaths and survivors of different natural disasters, spanning 150 countries. He then found that the amount of deaths seems to rise, the less democratic a state is. As failing in mitigating disasters can be considered a vulnerability of the state, it comes close to the core of this paper's subject. By using historical institutionalism, he explained the differences between the two main political regimes: democracies and autocracies. Similar to Chan (2013) did, he argues that democracies are inherently more resilient than autocracies are. He gives multiple reasons, of which the most important are that democracies have a better state capacity and are less
restrictive. This better state capacity means that the government is involved in more parts of society and has more resources already spread through it. This is because democracies have to consider the needs of their whole population instead of having a preference for just the political elites, as is the case with autocracies. By considering everybody, the state spends money on public health, education, and pensions. This social safety net makes sure that all citizens are less vulnerable to disaster, and these financial guarantees do not exist in most autocracies. The second reason, the restrictiveness of autocracies, means that they are less accountable and
political leaders censor this information. These restrictions are at the expense of the effectiveness and efficiency of their disaster mitigation. These effects would explain the fact that democracies are better in mitigating disasters than autocracies are.
That democracies are less vulnerable than autocracies is not necessarily a fact yet, as Boussalis (2011) examined how different economic, social, and political factors may increase damages and deaths in an earthquake. He used the U.S. Geological Survey database (PAGER), which includes data for all global earthquakes, and when analyzing this with the regime type, he did not find any evidence that it influences the mortality rate. This conclusion contradicts Lin's (2015) findings, which Boussalis explains by stating that in most democratic states, the electorate cares about the aftercare after a disaster, but is indifferent to preventative spending.
Pressure and Release model
For this thesis to be able to couple the regime type to vulnerability, a model is needed.
Vulnerabilities exist in every level of societies, while the political regime is present at the top. The Pressure and Release (PAR) model is a tool that can create a bridge between these different levels and help show which vulnerabilities link to the political regime, and why. The model's central concept is that a risk has two dimensions: the external hazard and internal vulnerabilities. The external hazard is, for example, a disaster that may strike a state, while the internal
vulnerabilities are weaknesses within its system. The summary of the model can is as the
equation below, which shows that if an actor scores high on both hazard and vulnerability, it has a higher risk (Wisner, Blaikie, Cannon & Davis, 2004).
The model splits this internal vulnerability up into three different layers. The first level consists of the root causes, which are ideas or systems deeply entrenched within the state. The primary examples are the political, demographic, and economic foundations or ideologies of a state. These causes influence its society everywhere: social and political systems, governmental decisions, and public policy. On this level, the (lack of) power of the authorities can be identified. This thesis will try to trace vulnerabilities to this level, as the political regime is always a root cause. The second level is about the dynamic pressures that can make a state more vulnerable. These are developments that are initiated at the highest level and create
vulnerabilities at the lowest level. An example is when a government focuses on major economic hubs (which would have an economic root cause), people can start to flock to these cities. The dynamic pressure that then forms is rapid urbanization, which will influence the lowest level of the model. This third level is coined the 'unsafe conditions' which are the situations citizens directly have to deal with in their lives. If we use the previous example of rapid urbanization, the cities that do not adapt need to deal with rising unemployment, which is such an unsafe condition as unemployed people do not have the resources to protect themselves. This means that when there are more unemployed people, that there is a higher risk of more fatalities when a disaster hits. This last level shows that vulnerability is not something that exists on a national or regional level, but even more on an individual level (Wisner, Blaikie, Cannon & Davis, 2004)(Nirupama, 2012)(Ciurean, Schröter & Glade, 2013, p. 10-11).
Regarding political regimes that influence the vulnerabilities, the authors of the model argue that these are the hardest to identify. Political leaders are hesitant to address or even acknowledge these vulnerabilities as they can negatively influence their power. Citizens may start to challenge their positions or demand structural changes within the state, which the political leadership wants to prevent. They do this by focusing on natural, technological, or other vulnerabilities, which are harder to control for them and, in turn, make them less accountable if their state is hit by a disaster (Wisner, Blaikie, Cannon & Davis, 2004).
The PAR model can examine a specific society and try to connect particular vulnerabilities to its higher causes. When we identify these vulnerabilities, it is easier to connect them to its cause and explain its importance in the whole equation of risk, hazard, and vulnerabilities. It has already proven its use as it is used in a plethora of studies, which focus on entirely different disasters. Tsasis & Nirupama (2008) used it to search for an explanation of how the spread of HIV/AIDS could develop as it did, and Rauken & Kelman (2010) used it to look for different river floods in Norway. In this thesis, the hazard is not necessarily taken into consideration as the focus lies on the vulnerability. It uses the model as a tool in the analysis to explain if the relevant
Conceptualization
This theoretical framework helps to develop a proper hypothesis based on our two main
concepts. These concepts, regime type, and vulnerability, first need to be conceptualized. While this is relatively straightforward for the concept of regime type, this can be quite complex for an idea as broad as vulnerability.
Political scientists often describe a regime type as the rules, norms, and procedures within the government and its officials. While these differences are mostly then categorized into groups as liberal democracies, communist regimes, or dictatorships, this categorization is dependent on what the exact subject of the research is (Reich, 2002). If the focus is on specific differences within the group of autocracies, it is better to split them up into their different variations. Autocracies can be divided into one-party states, dictatorships, military juntas, while democracies can be divided based on how many parties there are or on the extent of civil liberties. As this research focuses on the differences between democratic and non-democratic states, the regime type is solely based on a scale on how democratic a nation is.
The second concept is vulnerability, which can be interpreted in different ways. The narrow definition would mainly cover material damages or the number of people that have lost their lives due to disaster. This kind of narrow definition would need a disaster that we can generalize over all states, but there is no hazard included in this thesis. This means that these definitions are too narrow and not desirable for this thesis as the research needs to span most of the
vulnerabilities of a state. This research will then use a broader description, which will cover multiple economic, demographic, and political variables. It means that other variables are less likely to influence a possible correlation but also make it possible to generalize over all disasters instead of one specific event (e.g., an earthquake). For such a broad description, the concept of prevalent vulnerability will be used. This concept of vulnerability looks into all possible
weaknesses in a society that could lead to problems when it faces a crisis. In short, it focuses on the overview of ‘favorable conditions for direct physical impacts, as well as indirect and, at
Hypothesis
If we connect all theories and examples from the theoretical framework, some regularities show up. Chan (2013) found that the institutional differences between democracies and autocracies have a significant impact on the whole of disaster management, and Lin (2015) connected these differences directly to factors that would make autocracies more vulnerable by using historical institutionalism. Other research showed that a multitude of different variables also influences the vulnerability of a state, like its GDP, and how well a state is connected to the rest of the world.
The regime type of a state could influence many of these variables, as it consists of the complete set of norms, rules, laws, and uses of the state and its officials (Reich, 2002). The Pressure and Release-framework show that this, the core of the state, should be identified as one of the root causes of vulnerability that underlie dynamic processes and unsafe conditions. By this logic, a correlation between the regime type of a state and how vulnerable it is to disasters seems
plausible. If this is the case, we can regard the regime type as more of an indicator than a cause; some processes within a particular regime are the cause, not the type itself.
To suggest the direction of the correlation, we use the findings of Lin (2015) and Chan (2013), who found that democracies are more capable of handling disasters than autocracies, which suggests that they may be less vulnerable. Persson & Povitkina (2017) also investigated how democratic institutions might influence disaster preparedness in the case of floods and found that the more democratic a nation is, the better it is prepared for disasters. As preparedness is the mitigation of vulnerabilities in the institutions, infrastructure of socio-economics of the state, it also suggests that democracy makes states less vulnerable.
Academics seem to agree that democracies should be and generally are less vulnerable. This assumption, however, does not stem from focused research, but from papers which looked into specific types of disasters. Due to this fact, the arguments of these papers are too specific to use for this study, but it gives us some ideas about what dynamics would make a specific regime type more or less vulnerable to crises. Plümper & Neumayer (2009), for example, found that
democracies and autocracies deal differently regarding famines: the population as a whole has more influence in their democratic government, and thus this government will act with policies
that benefit the whole population. In autocracies, a small elite has more influence, and they instead would choose to compensate sparingly. In their studies, this explains why famines are worse in autocracies, but the underlying dynamic could work similarly in other crises. Another systematic difference that is often found which relates closely to this example is that the political figures in democracies want to retain their political careers. This means that it is in their best interest to help people, and as their support bases consist of a big part of the population, it is hard to exclude certain groups without losing support. This incentive does not exist for autocracies, and they see disaster preparedness or assistance as something that uses up using costly resources. The leadership would instead use these to keep the people they need to stay in power happy. For autocracies, it is simply not efficient to help everybody (Flores & Smith, 2013).
Considering that for specific disasters, democracies are considered less vulnerable by academics and that others have identified processes that could explain why this is, the hypothesis will be as shown below.
H1: The less democratic a state is, the more vulnerable it is to disasters
As stated before, this hypothesis is expected as many papers of academics point towards this conclusion. Still, this has not been researched for vulnerability in its broadest meaning but mostly for when a specific disaster hits them. By using the concept of vulnerability in the broad sense of how well states are prepared for any disaster, the conclusions can be more easily generalized than research that focuses on specific hazards. Also, this paper breaks with most literature by selecting states based on how democratic they are instead of by which disasters they are hit. This selection also increases how well it can be generalized and helps to try to confirm what is heavily suspected but has not been proven irrefutably: that democracies are always less vulnerable to disasters.
If it seems that there is no correlation present between the regime type and vulnerability, the following null-hypothesis is true:
H0: The regime type of a state does not influence its vulnerability to disasters
If the null hypothesis turns out to be accurate, there would be many consequences. The expectations made by current literature would seem not to be accurate, which means
explanations need to be sought to understand this indifference in the vulnerability of regime types. It could mean that there are specific disasters in which the regime type can explain the state's vulnerability, but this is not the case for most disasters. It could also mean that economic indicators are just far more important than the political indicators, and other researchers have mostly focused on wealthier, democratic nations. In any case, it would open up many questions for future research.
Chapter 2: Methodology
This chapter will explain in detail how to replicate this research and it explains why the research design is as it is. The first part will shortly explain the main design, after which the concepts which have been discussed in the theoretical framework are operationalized. The next step is setting the rules for the case selection and using these to select the twelve cases that will be examined. The last section explains how the data for these cases is collected and how this data will be processed. The complete methodology will enable this research to answer the question ‘To what extent does the type of regime affect its disaster vulnerability?’
Research design
This research is empirical and explanatory research with the regime type as the independent variable and vulnerability as the dependent variable. As the primary goal of this research is to find differences between different regime types, multiple cases need to be analyzed. This means a quantitative, comparative case study will be used to determine the relationship between the two variables. This design makes it possible to analyze a multitude of countries that have different regime types and compare the outcomes (Bryman, 2012, pp.72-76).
First, we select twelve states which are going to be analyzed in this paper. This selection is based on a most different system design, with some additional criteria which will be presented in the case selection-section. The selected countries are Australia, Brazil, Botswana, Cameroon, China, Denmark, Lebanon, Malaysia, Mauritius, Nicaragua, Sudan, and Ukraine. These will be divided into four categories of political regimes based on how democratic they are. The categories are the democratic nations, flawed democracies, hybrid regimes, and authoritarian regimes, and these represent a scale from most democratic to least democratic regimes. States being more
democratic means that the state institutions have a more robust democratic groundwork and that these democratic values penetrated more of the state's society. Each of these four categories will include three countries, which means a total of twelve states will be analyzed. Using the most different system design decreases the chance that there are variables other than regime type within all cases, which have an apparent effect on its vulnerability as this could (at least partly) falsify the results. The reliability of this research is harder to safeguard, mainly as only a dozen
that if others replicate this research but use other cases, the results will be valid but may differ from the findings of this research. It also means that the initial results may be harder to
generalize, but it still is able to give a good indication of the relationship (Bryman, 2012).
Next to the regime type, a score for vulnerability will be assigned to these twelve states. The Prevalent Vulnerability Index will be used, which means that multiple numerical variables are assigned to each case. This index will be used to find this possible relationship and will be analyzed with the earlier described Pressure-and-Release model as a framework. With this framework, the practical feasibility of this research is high, with just the main challenge of finding all the correct data for one of the concepts.
Operationalization
We will start with operationalizing the two main concepts, as the case selection depends heavily on how the political regime is measured. Two already existing indexes will be used: the regime type will be measured based on the scores on the Democracy Index. Cordona (2008, pp. 8-11) developed a framework that can help operationalizing vulnerability, as his Prevalent
Vulnerability Index measures it in its broadest sense.
The regime type is the independent variable, which should influence the vulnerability of a state. To make the type of regime measurable, this paper uses the Democracy Index of the Economist Intelligence Unit (2015). The Democracy Index is a score of 1 to 10, which is based on 60 indicators spread over five main categories: electoral process and pluralism, civil liberties, government functioning, political participation, and political culture. The classification of the same report will be used to determine which states belong to which political regime. The highest scores of 8 to 10 will classify it as a full democracy, 6 to 8 as a flawed democracy, 4 to 6 as a hybrid regime, and less than 4 as an authoritarian regime (The Economist Intelligence Unit, 2019, pp. 51-52).
In their terminology, democratic states meet specific minimum requirements of political freedoms and civil liberties and have a culture that leads to the flourishing of the state and its democracy. In a flawed democracy, this last aspect is not present, and there are a few weaknesses that can be a danger to democracy. Their institutions are mostly stable, but democratic ideals are not entrenched in society. After the flawed democracies, there are hybrid regimes that mostly have elections, but these are not fair or free. In these states, weaknesses in democratic institutions are widespread, and their civil society is weak or nonexistent. The last group is composed of authoritarian countries, in which one actor rules the state, and there is little room for opposition or change of the political elite. If they even have any democratic institutions, their substance is minimal (The Economist Intelligence Unit, 2019, p. 53). By using this index, precise lines for regime types are established, which are both needed for case selection and effective analysis of all the data.
The dependent variable, vulnerability, will be measured using the Prevalent Vulnerability Index as described by Cardona (2005, p. 8-11). This index covers a wide array of parameters over three categories: Exposure, fragility, and lack of resilience. The variables in the first category,
exposure, shows how susceptible the state is to external hazards. The assumption is that the variables in this category are generally good to determine if a nation is at risk, which means it works for every disaster. The second category is fragility, which measures the possible socio-economic effects on the state's population and is based on an intrinsic belief that there are adverse effects on social structures when disaster strikes. The last category, lack of resilience, is built around measuring how well a society can absorb or recover from a disaster. In the original index, these scores are inverted and hence a lack of resilience. One significant change made in this research is that we do not invert the scores, subtracting them from the other indicators. This change is explained in more detail in the section about data analysis.
The original index was published more than ten years ago, which means that some variables were unusable. In some cases, the databases do not have the required data anymore, while in others, the idea behind certain variables is outdated. To make sure the index is still useful, we need to make a few changes. Firstly, two indices have been updated by their managing
and the Environmental Sustainability Index has become the Environmental Performance Index. In ‘resilience’, social expenditure is changed as it would generally be the sum of spending on healthcare, education, and pensions. However, the data on pensions was not available in most countries, so it was dropped from the equation. Some other variables needed to be replaced altogether. In the exposure category, the total capital stock of a state needed to be known. The idea was that this would show how much capital could be damaged by a disaster. This is replaced with an infrastructure score with logic that the more (advanced) infrastructure a state has, the more of it can be lost. The last variable of the fragility indicator measured the soil quality using the GLASOD database, but it has not been updated for over two decades. Its scores were also region-specific instead of per country, so it is swapped for a new indicator that measures the average rate of erosion per year. In the last category, resilience, two variables also have been replaced. The first one is the number of televisions per 1000 inhabitants, meant to measure the spread of information throughout the country. As nowadays, mobile phones are a better indicator of this phenomenon, it is replaced by how many mobile cellular subscriptions there are per 1000 people. The other variable that has been changed was which percentage of the entire
infrastructure of a state was insured, but as this data was not accessible, it has been replaced by the total insurance premium volume as a percentage of the GDP. The updated index, including all 24 variables, is shown below.
Figure 1: The Prevalent Vulnerability Index
Exposure Fragility Resilience
Population growth Global Multidimensional Poverty Index* Human Development Index
Urban growth Dependents Gender-related Development Index
Population density Social disparity Social expenditure*
Poverty-population below US$ 1 per day Unemployment Governance Index
Infrastructure score* Inflation Total insurance premium volume
Imports and exports of goods and services Dependency of GDP growth of agriculture Mobile cellular subscriptions per 1000 people*
Gross domestic fixed investment Debt servicing Hospital beds per 1000 people Arable land and permanent crops Average rate of erosion* Environmental Performance Index* * These variables have been updated and are different from the variables in the original Prevalent Vulnerability Index
Next to changes in variables, this paper made some changes in data collection and analysis. Cardona (2005, p. 115-116) describes that for the Prevalent Vulnerability Index to work best, data needs to be collected over multiple years. For this research, we will only focus on the year 2015 as the time to conduct this research is limited, and it will provide a clear view of
differences between countries. There are two reasons for choosing 2015: firstly, it lies a few years in the past. This makes it plausible that the data that is needed is already available, as it can take a few years before it is published. Secondly, it is not too far in the past, which could mean the conclusions of this research would immediately be outdated.
Cardona (2005, pp. 58-60) lists where the data for these variables can be found, which is mostly in easily accessible, open databases by (inter)national institutions. The central database, ‘World Development Indicators’, of the World Bank, is used for most of the variables. Only when data is not available in this or other databases of the organization, other sources will be used and
mentioned as notes (World Bank, 2015). As stated before, the most significant benefit of this database is it fully accessible online and has an incredible amount of data for all states. In most cases, the database also gives information on where they got the information from which
increases reliability. One downside which Cardona (2005) also acknowledges is that the database is far from complete and that sometimes the data needs to be filled in otherwise. Preferably, data that lies closest to the year 2015 is used, and if this were not possible, an estimation would be made. This estimation will be mentioned as a note in the appendix for full disclosure.
By using this updated index, it is possible to give every state a score that fits their vulnerability level. While the states may exist in entirely different realities, it enables proper comparative research to search for significant similarities and differences. As the full Prevalent Vulnerability Index exists of three sub-indices, it is also possible to look for differences within exposure, fragility, and resilience. To see the full index and how every variable is measured, it is included as appendix 1.
Case selection
As we want to know to see how the political regime influences their vulnerability, the actors that will be our cases need to be countries. The whole population of these consists of all recognized states by the United Nations, currently around 193 (Green, 2018). Twelve of these will be
selected randomly, but they do have to meet a few criteria. Firstly, the state needs to have a score assigned in the Democracy Index of 2015 (The Economist Intelligence Unit, 2015), as this determines regime type. Secondly, as there are four categories of regimes, every category needs to have three states assigned to create equal groups. The third criterion is that states assigned within the same regime type can not be located on the same continent. This division strengthens most different system designs as possible regional effects than are excluded. Lastly, different studies show that small (island) states are harder to generalize as they have some unique characteristics which influence their situation (Pelling, & Uitto, 2001) and seem to be more likely to be democratic (Anckar, 2002). To exclude these, the population of a state needs to be at least 1 million.
The following twelve states have been selected based on these requirements that are spread out globally and differ widely in GDP. This way, the design respects the most different system design.
Figure 2: The selected cases
Democracies Flawed democracies Hybrid regimes Autocracies
Denmark Botswana Ukraine Cameroon
Australia Brazil Nicaragua China
Data analysis
When all the data is collected, it needs to be converted into the indicators. Cordona (2008) has one preferred method which includes using multiple years, as stated before, and adding weights to each value. He acknowledges that sometimes other methods are preferred and also explains these in detail. Therefore, this thesis will weigh every variable equally, as the process of finding and fairly weighing them according to the original method is very complicated and
time-consuming. As we only work with one year and equally weighted variables, translating the data in the indicators also needs to be changed. All variables will be converted into Z-scores, and the sum of 8 variables will create one of the three sub-indicators. Z-scores show how many standard deviations the score of a variable differs from the average score, giving a clear overview of how these states compare to each other. This standardization makes sure that variables with outliers have less impact on the indicators, which is a benefit. On the other hand, the standardization also means that when the data of a variable is spread further from the mean, this variable will weigh slightly more when calculating the indicators.
These Z-scores, however, will give a clear overview of how states compare to each other. The variables will then create the three indicators: one which measures the exposure (PVI-ES), one which measures fragility (PVI-SF), and one which measures resilience (PVI-LR). The final indicator, PVI, is created by adding up the first two indicators and subtracting the last one. In Cardona's original work (2005), the last indicator was inverted, and all three indicators were added up. Instead of inverting the last sub-indicator, our analysis will subtract it from the other two. This is done as the two first sub-indicators measure shortcomings within the state and the last one their strengths. In the end, this still means that a higher PVI-score means that the state is more vulnerable to disasters than states with a lower score.
When all data is collected, the analysis will start with quantitative data analysis in which the relationship between the regime type and PVI (and the three categories it consists of) is determined. To do this, we can use a simple regression and a Pearson correlation. The simple regression shows how an increase/decrease in the Democracy Index may influence the score in the PVI, while the Pearson correlation can determine the strength of the correlation between the two variables (Vocht, 2014). While these quickly can show if a relationship is present, it limits the research as it only measures for linear relationships. This limitation means that if hybrid regimes or flawed democracies are more vulnerable than democracies or autocracies, the analysis would not show it but would show that there is no relationship at all. This limitation is
considered, but as the theory suggests that the relationship would likely be linear, these analyses are still the most useful.
After this data has been processed, the outcomes will be analyzed with the Pressure and Release-model (Nirupama, 2012) as a framework. This analysis will be used to explain the (lack of) relationship we found earlier in the analysis, and can explain for example how some indicators may be inherently weaker in autocracies or find more general commonalities within regime types and differences between them.
Chapter 3: Analysis
Now that the theoretical framework and methodology are established, all the tools that are needed to answer the question if the political regime influences the vulnerability of the state are in place. The theoretical framework suggests that it is the case that democracies are less
vulnerable, as other researchers found this by looking into the consequences of specific disasters in specific regimes and argued that democratic institutions are more robust than other regimes. In the methodology, this thesis built a research design with which twelve states in four categories of political regimes can be compared based on how vulnerable they are. The result and analysis of these results will be discussed in this chapter.
Empirical results
Before we analyze the data with the help of statistics, we look at the gathered data. This data is available from appendix II to VII, which are the tables with the Democracy Index and all variables for the Prevalent Vulnerability Index. The data is sorted based on their score in the Democracy Index, going from most democratic to least democratic, and a quick overview of the indicators is given by figure 3 at the end of this section.
The scores of the Democracy Index seem to be distributed normally over the 12 states, as in most cases, the score lowers with approximately 0,4 to 1 point. This range can be expected as the difference between the highest and lowest score is 6,74. When we divide this by the number of cases, we find that the average difference is 0,61, which fits these scores. The only two
exceptions are Denmark and Australia, as between them, the difference is just 0,11, and between Lebanon and Cameroon, which is a bit more than 1. We can explain the small difference between the two democracies by the fact that there are fewer democracies than in other categories, and these democracies are clustered geographically. The whole group of democracies consists of 20 states, of which 14 are European. As there could not be two states from the same continent in the same group of political regimes, the combination significantly increased the chance that the Democracy Index would be closer to each other when two states would be selected with the criteria of this paper. The difference between Lebanon and Cameroon does not have a clear identifiable cause and is considered a natural outlier.
Next, we will look at data that is collected to fill the four indicators. The first one, PVI-ES, shows how exposed a state is towards a possible hazard. The variables, as seen in appendix I, are mostly demographic and economical of nature. When the data is examined, it becomes clear that none of the variables follows a clear pattern. This pattern is naturally the same for the combined indicator, and this can be illustrated by the scores of the two most democratic states. Denmark has a score of 1,40, which makes the state one of the more exposed states, while Australia has a score of -1,5, which is below average. The next indicator, PVI-SF, measures the fragility of the cases. These variables are also mainly economic, but this time their data looks to be distributed more orderly. For all variables, democracies score lower than all the other regimes, while most authoritarian states have a higher score. One pattern that emerges, however, is that the scores for flawed democracies and hybrid regimes are distributed more randomly, which becomes apparent when the indicator is examined. The two top democracies score the lowest, two of the
autocracies score the highest, but there is no clear pattern for all cases in between. The last sub-indicator, PVI-LR, measures how resilient a state is, and other than with the other two indicators, a higher score has a positive effect on vulnerability. For these variables, the pattern of data is evident over all of the eight variables: the less democratic, the lower the score. As with the second sub-indicator, this is most clear with the most and least democratic nations, but in this case, the data is less scattered within the groups of flawed democracies and hybrid regimes. The only apparent exception in this data is that China has a higher score than would be expected.
Lastly, the PVI indicator will be explored, a combination of the three earlier discussed sub-indicators. There is a general pattern between the Democracy Index and this indicator as to the score increases when looking from the most democratic nation to the least democratic one. There are exceptions like Mauritius and China, but for the most part, the data seems to suggest that the political regime does influence the vulnerability of a state. A few statistical tests need to be passed to confirm this, which the following sections will do.
Figure 3: The Prevalent Vulnerability Index per country
Category Country DI PVI-ES PVI-SF PVI-LR PVI
Full Democracy Denmark 9,11 1,40 -5,05 9,88 -13,53
Australia 9,01 -1,57 -4,43 6,69 -12,69
Mauritius 8,28 0,46 4,05 2,10 2,41
Flawed Democracy Botswana 7,87 1,37 2,47 3,22 0,62
Brazil 6,96 -4,46 -0,73 1,55 -6,74
Malaysia 6,43 1,14 -3,80 2,10 -4,76
Hybrid regime Ukraine 5,70 -1,90 1,69 3,86 -4,07
Nicaragua 5,26 -0,50 1,45 -2,04 2,99
Lebanon 4,86 3,47 -1,69 -4,69 6,47
Autocracy Cameroon 3,66 1.66 2,84 -8,32 12,82
China 3,14 0.92 -3,45 -1,86 -0,67
Test of Normality
The last table, figure 3, shows a summary of what the states scored for the Democracy Index, the PVI and the sub-indicators of the PVI. While the first empirical results suggest that there is a relation, we need to know if the relationship is significant. We will later test for this correlation, but to be able to do this, we need to make sure that the PVI is distributed generally over the different regime types. If it is normally distributed, it means that data is more frequent at the mean and that there is less when moving from it. We expect this to be the case for the whole population of states, but does our small sample reflect this distribution? This distribution is what the Shapiro-Wilk test measures, and this is needed to do other statistical tests like the Pearson correlation correctly.
This test will look at how similar the observed and normal distribution are and then calculate if the difference is significant enough to reject the idea that it is normally distributed. The data from figure 3 is used in SPSS to run the test, and figure 4 shows the test results. The most important thing is that when the level of significance is below 0,05, the null hypothesis is rejected, which means that the data would not be normally distributed. For all our political regimes, the score is higher than 0,05, which means we can reject this null hypothesis. In this case, this means that the data for the PVI is normally distributed for all regime types.
Figure 4: The Shapiro-Wilk test
Regime type Statistic df Sig.
Democratic 0,789 3 0,089
Flawed dem. 0,934 3 0,502
Hybrid reg. 0,963 3 0,630
Pearson Correlation
The empirical results seem to point towards a relationship between the Democracy Index and the Prevalent Vulnerability Index. To ensure this is the case, we need to test that the correlation exists and that it is significant. If two variables correlate, it means that the one variable has a relationship with the other, and then we can measure how much they influence each other Its significance is important because it shows if a relationship exists at all. Shortly, when the correlation is too weak or turns out to be insignificant, we can not conclude that there is a relationship between the DI and the PVI or its sub-indicators.
We used SPSS to calculate a Pearson r correlation coefficient, to test if this relationship is
present. The outcome of this bivariate correlation can be found in figure 5 and shows that there is a strong correlation between the score given for the Democracy Index and the PVI. The result implies that there is a strong negative correlation between the vulnerability of a state and how (non-)democratic this state is, as Pearson r is -0,77. Its significance is also below the 0,01 threshold, which means it is significant and that our earlier observations are not caused by, for example, a random sampling error.
Figure 5: The Pearson Correlation for DI & PVI
DI PVI
DI Pearson Correlation 1 -0,777**
Sig. (2-tailed) 0,003
N 12 12
When we split the PVI indicator into its three components and the analysis is done again, a few apparent differences between these sub-indicators appear. The indicator which measures exposure (PVI-ES) has a very weak negative correlation, which is not significant. This finding corresponds with the empirical results, where the data seemed mostly randomly distributed and not tied to the Democracy Index. The second indicator (PVI-SF) has a far stronger negative correlation but is not significant, which is not surprising. The last sub-indicator (PVI-LR) differs from the others as it has a significant and strong positive correlation. While expected, it shows that this sub-indicator has a significant influence on the two others and, more importantly, the final PVI-indicator. A summary of these findings can be found in figure 6.
Figure 6: The Pearson Correlation for DI and the three sub-indicators
DI PVI-ES PVI-SF PV-LR
DI Pearson Correlation 1 -0,052 -0,405 0.888**
Sig. (2-tailed) 0,872 0,192 0.000
N 12 12 12 12
**Correlation is significant at the 0.01 level (2-tailed).
The Pearson Correlation shows that the combined set of indicators can say something about a possible relationship between the independent and dependent variables, while the three sub-indicators on their own are not always influenced by how democratic a state is. What this means will be explored after we double-check these findings by using a linear regression.
Linear regression
As the relationship between the DI and PVI is continuous and linear, we can use the linear regression to predict the value of the Prevalent Vulnerability Index by using the Democracy Index. These results may create more understanding of how the relationship works and how strong it is by giving us more concrete numbers while confirming the findings of the Pearson correlation.
When we input the PVI-indicator together with the Democracy Index into the linear regression function of SPSS, we find that that the former seems to be significantly influenced by the latter. It also shows that the regression can explain 60.4 % of the variation within the data. This means that 60.4% of the data falls within a reach that is expected by the model and can be explained by the linear regression. The coefficients also predict that when the Democracy Index rises with one point, the PVI indicator shrinks with -3,157 on average. If a state would have a Democracy Index of 0, it predicts that the PVI score would be 19,113. On figure 7, the black line represents this equation and the scores of the countries on both the DI as the PVI.
When the PVI indicator is split into its sub-indicators, it becomes clear that the results match these of the Pearson Correlation. It shows that, with a significance of 0,694 (> P = 0,05) and an effect of 0.067 that the Democracy Index does not influence the PVI-ES indicator. The second sub-indicator, PVI-SF, is still insignificant with a significance of 0,218 (> P = 0,05) But, it would have been more influenced by the Democracy Index with a score of 0,160. Mainly the variation in the data creates the situation in which the trend is not significant. PVI-LR, however, has a significance of 0,001 (< P = 0,05) and an effect of 0.383, which means that the Democracy Index influences it. The scatterplots that show these lines, as shown in figure 7, are available in the appendix. These findings confirm the relationships found with the Pearson Correlation and give them insight into how strong they are. It still leaves the question of why these two models find a negative correlation for the final indicator, but that there are significant differences between the sub-indicators. These differences will be explored in the next section of this chapter.
Statistical analysis
As stated before, while Democracy Index does have a significant influence on the Prevalent Vulnerability Index, its subsequent sub-indicators do not. The main reason that there are
differences between the sub-indicators comes from the fact that every sub-indicator uses a set of variables that focus on a different aspect of a country. While politics easily influence some of these aspects, others are hardly influenced by them. This section aims to identify these
differences, explain why the political regime more easily influences some, and find out how this impacts the rest of the analysis.
The first sub-indicator, PVI-ES, measures exposure by using different variables that mainly focus on demographical characteristics as population growth, urban growth, and how many people live in poverty. It is hard to tie the regime type's direct effect on most of these variables. While it could be that one regime type may be more effective in eradicating poverty and scoring lower on that variable, most of these variables are hard to influence politically. One clear
example of this is how the third variable, population density (ES3), naturally varies between all regime types. It is hard to think that national policies can have a significant effect on this
variable. Without such an effect, it means the data is more randomly distributed. The data shows that Denmark and Australia, which are both democratic nations, have entirely different
population densities. China, which is one of the more autocratic nations, is far more comparable to Denmark. These variables are the majority in the indicator PVI-ES, which can explain why the political regime does not influence its vulnerability.
The second sub-indicator, PVI-SF, has a stronger relationship with the Democracy Index, but it is still insignificant. The variables which it consists of are not demographical of nature, but mainly economical. They range from their yearly inflation rate, to unemployment, to how wealth is divided throughout the nation. This stronger relationship shows that these economic variables are more influenced by political decisions than demographics are, but not why the relationship is still not significant enough. This difference, in turn, can be explained by that economics and politics are not as connected as is often thought. The idea that many democracies are more prosperous than average is not necessarily wrong, but the variation between the states has
Inklaar & Timmer, 2015). A variable as the Global Multidimensional Poverty Index (SF1) is still heavily influenced by political regimes as it also has a social aspect: The scores for this socio-economic variable rise when a state becomes more authoritarian. For other variables that are purely economic, however, there seems to be far less of an effect. The lack of this effect can be seen, for example, if the scores of social disparity (SF3) and inflation (SF5) are compared between the different regime types.
Other than those first two, the last sub-indicator, PVI-LR, did have a significant and strong correlation. This indicator was not created from demographical or economic variables but mostly from social ones. The political regime easily influences these variables as these are mostly about the freedoms and treatment of their inhabitants. For example, the Human Development Index shows a clear pattern in which more democratic states scored better. This pattern is also true for social expenditure, the insurance premium volume, and the Sustainability Index. It shows that these social variables built resilience throughout the state, are heavily influenced by the political regime. As this is effect is felt over almost all variables in the indicator, the whole indicator has a clear relationship with the Democracy Index. It tells us that the more democratic a regime is, the more it invests in protecting its citizens.
How can it be that while only one of the three sub-indicators has a significant correlation with the Democracy Index, that the whole indicator PVI is still significant? This third indicator's effect is mainly this strong because the data from the first sub-indicator is completely scattered. This randomness means that when the second sub-indicator is added, a minor pattern emerges. This pattern, however, is not significant enough to get a relationship between the Democracy Index and the combined PVI-ES and PVI-SF-scores. However, the relationship between the political regime and resilience (PVI-LR) is strong enough to offset the randomness created by the first two. This means that while keeping the unpredictable exposure and fragility of a state in mind, how resilience a state is can be predicted by its regime type and, therefore, the
